Lightweight Real-time Spatial Audio Interpolation for Standalone VR using Hand Claps

Kim M. Lightweight Real-time Spatial Audio Interpolation for Standalone VR using Hand Claps. In: Express Paper. Audio Engineering Society; 2026.pp. 448:1-. Available from: https://aes.org/publications/elibrary-page/?id=23167

@inproceedings{Kim2026_23167,
  author = {Kim, Minsu},
  title = {{Lightweight Real-time Spatial Audio Interpolation for Standalone VR using Hand Claps}},
  pages = {448:1-},
  year = {2026},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=23167}
}

TY  - CPAPER
TI  - Lightweight Real-time Spatial Audio Interpolation for Standalone VR using Hand Claps
AU  - Kim, Minsu
SP  - 1
M1  - Paper 448
PY  - 2026
DA  - 2026/05/28
UR  - https://aes.org/publications/elibrary-page/?id=23167
PB  - Audio Engineering Society
LA  - en
AB  - This paper proposes a lightweight room-adaptive reverberation method for standalone VR using sparse in-situ impulse measurements. Rather than reconstructing a full acoustic field, the proposed system estimates a small set of reverberation-related descriptors, including RT60, total reverberant level, early reflection level, and late reverberation level, and maps them to an algorithmic reverberation backend. The descriptors are interpolated over source and listener positions using a metric-specific double inverse-distance-weighting (IDW) formulation. The interpolation power is estimated separately for each descriptor from additional tuning measurements, allowing the system to adapt to the spatial behavior of each acoustic parameter. The method is first evaluated on two public room impulse response datasets, dEchorate and Robots Journey, to examine whether sparse measurements can support descriptor interpolation under different reverberation conditions. In both datasets, the proposed p-estimation method performs closer to an oracle upper bound than to a single measured-value baseline. The method is then applied to a real acoustically non-uniform room measured using smartphone microphone and hand-clap impulses. Objective evaluation shows that multi-anchor conditions substantially reduce late-reverberation error compared with a 1-anchor baseline, while 5-anchor and 9-anchor conditions produce similar accuracy. A pilot listening test with five participants further indicates that the 5-anchor and 9-anchor conditions are judged as more realistic than sparser conditions. These results suggest that sparse in-situ measurements can provide a practical correction layer for room-adaptive VR reverberation.
ER  -